|MENTASTI SIMONE||Cycle: XXXIII |
Section: Computer Science and Engineering
Tutor: AMIGONI FRANCESCO
Advisor: MATTEUCCI MATTEO Major Research topic
:Embedded AI and Sensor Fusion for Autonomous VehiclesAbstract:
One of the main requirements of an autonomous vehicle is the ability to sense the area around the ego-vehicle and retrieve a uniform representation of the surrounding. A standard setup for a self-driving car consists of multiple sensors (i.e., Lidar, radar, cameras, IMUs, Encoders). The goal of sensor fusion is to collect the heterogeneous information provided and merge them, to give the control algorithm a fully characterized but concise representation. Each sensor offers a specific feature of the surrounding obstacles; cameras are used for classification, while Lidar provides accurate position and radar relative speed. Aggregating all those information and considering the position of the obstacle (i.e., on the street, on the ego-vehicle trajectory, parked on the roadside) requires accurate synchronization between the sensors and high computational power to perform fusion in real-time. My research focuses then on realizing a sensor fusion pipeline on a development platform equipped with limited computational power and soft synchronization between only some sensors. The fusion process will then be performed asynchronously, employing less traditional data representation, like a 2D occupancy grid, generally used in robotics systems, characterized by limited computational power, but not in the autonomous driving field.